The official getuserfeedback.com Agent Plugin brings product feedback into the agent workflow. It connects to the hosted getuserfeedback.com MCP server and adds focused skills for collecting feedback, reviewing responses, and following up with users.
Use the route for your client. The public product name is always
getuserfeedback.com; getuserfeedback is the shared technical slug used by
plugin hosts and may appear in install commands or namespaces. Agent Plugins
standardizes the package, but each client still owns installation and
authentication.
Repository: https://github.com/getuserfeedback/agent-plugin
Add the repository as a marketplace, then install the shared technical slug:
/plugin marketplace add getuserfeedback/agent-plugin
/plugin install getuserfeedback@getuserfeedback
See the Claude Code plugin guide.
Open the Command Palette, run Chat: Install Plugin From Source, and enter:
https://github.com/getuserfeedback/agent-plugin
Install directly from GitHub:
copilot plugin install getuserfeedback/agent-pluginSee the Copilot CLI plugin reference.
Open Powers → Add Custom Power → Import power from GitHub, enter the repository URL, and choose Install. See Install powers.
Cursor supports this as a local import. Clone or copy the repository to
~/.cursor/plugins/local/getuserfeedback, then restart Cursor or run
Developer: Reload Window. See Test plugins locally.
On Cursor Teams or Enterprise, an administrator must first enable Allow Local Plugin Imports under Dashboard → Settings → Security & Identity → Marketplace and Plugins. The setting is off by default on Enterprise.
mkdir -p "$HOME/.cursor/plugins/local"
git clone https://github.com/getuserfeedback/agent-plugin "$HOME/.cursor/plugins/local/getuserfeedback"Add the repository as a marketplace:
codex plugin marketplace add getuserfeedback/agent-pluginIn Codex CLI, enter /plugins, choose the getuserfeedback.com marketplace,
install the plugin, and start a new session.
In the ChatGPT desktop app, restart the app after adding the marketplace, open
the Plugins Directory, choose the getuserfeedback.com marketplace, and
install the plugin. Each user completes OAuth for their own getuserfeedback.com
account; the package contains the registered MCP app identifier, not user
credentials.
See OpenAI's plugin packaging guide and the Codex CLI plugin browser.
Install from GitHub, trust the plugin's skills and MCP configuration, then enable it:
grok plugin install getuserfeedback/agent-plugin --trust
grok plugin enable getuserfeedbackStart a new session after installation. See the Grok plugin guide.
Hermes, OpenClaw, and NanoClaw can load the package, but cannot currently complete the OAuth flow required by the hosted getuserfeedback.com MCP server from this package's metadata. The skills may appear, but the MCP tools will not be usable. These are not full installation paths yet.
Hermes does not currently expose portable MCP servers to its OAuth login command. The affected package flow is:
hermes plugins install getuserfeedback/agent-plugin --no-enable
hermes plugins enable getuserfeedbackTrack the Hermes portable OAuth issue.
For OpenClaw, the bundle can be installed from a local checkout, but its native
auth: "oauth" setting is not part of the portable package:
git clone https://github.com/getuserfeedback/agent-plugin ./getuserfeedback-agent-plugin
openclaw plugins install ./getuserfeedback-agent-plugin
openclaw plugins inspect getuserfeedbackSee the OpenClaw bundle guide and native MCP OAuth guide.
NanoClaw accepts Agent Plugins only from its local templates directory. Its credentials proxy does not provide the MCP OAuth flow required here:
git clone https://github.com/getuserfeedback/agent-plugin /path/to/nanoclaw/templates/getuserfeedback
ncl groups create --template getuserfeedback --name "getuserfeedback.com"See NanoClaw's agent template guide.
The plugin connects to https://mcp.getuserfeedback.com/ over Streamable HTTP.
On clients that support OAuth for plugin-declared HTTP MCP servers,
authentication uses the client's OAuth flow. The plugin contains no credentials
or fixed authorization headers.
This plugin is distributed under the MIT License. See LICENSE for the standard license text and copyright notice.
collect-product-feedbackcreates and safely updates focused surveys.review-user-feedbackreads responses, inbox items, and weekly digests.follow-up-with-usersdrafts and sends approved respondent follow-ups.
The package includes native Claude Code metadata alongside the portable Agent
Plugins v1 manifest. Every variant uses getuserfeedback as its technical slug
and presents getuserfeedback.com as the product name wherever the host
supports separate display metadata.
For help, see getuserfeedback.com support.
Outcome:
- Maintainers can safely turn an approved stable plugin version on public
maininto an installable, byte-verifiable release without cross-repository credentials or manual repair of matching draft state.
Promised behavior under named conditions:
- Across reruns of the original failed workflow run and GitHub read-after-write delay, a matching draft is resumed. Missing assets, interrupted uploads, and expected-name assets with mismatched bytes are deleted or uploaded until the draft matches. A matching published release is verified and accepted.
- Across overlapping runs, SHA-scoped concurrency does not cancel another commit's run, and reconciliation remains scoped to the version-derived tag.
- Across every retry, the release must point to the exact triggering commit,
use the canonical
getuserfeedback.comtitle and version-derived artifact names, and contain the exact archive and checksum bytes produced from that commit. The workflow fails closed instead of replacing mismatched published state.
Accepted failure or uncertainty:
- A commit can be present on public
mainbefore its release is available. - A GitHub outage or visibility delay can leave a matching draft that requires rerunning the original failed workflow run at its original commit.
- Recovery ends when GitHub no longer permits another rerun of the original workflow run or its retry allowance is exhausted.
- Mutated release identity, title, tag target, unexpected asset names, and any mismatched published state are unsupported and refused rather than repaired.
- Mirroring source into the public repository is a separate boundary; this workflow does not provide a cross-repository transaction.
- Archive production uses the selected GitHub runner's system tools. Toolchain changes can produce different bytes and cause a later retry to fail closed.
- Latest-release selection depends on GitHub's legacy SemVer behavior.
Evidence:
- The workflow byte-compares both downloaded assets before and after publication.
- Contract tests in the canonical source repository cover planner decisions for initial publication, interrupted-upload and mismatched-byte reconciliation, exact release identity, and mismatched-state refusal. A shell harness exercises the workflow's bounded visibility-delay polling behavior.
- Contract tests lock the SHA-scoped, non-cancelling concurrency configuration and delegation of latest-release selection to GitHub's legacy SemVer mode.
Escalation triggers:
- Revisit this design if releases must recover without rerunning the original workflow or beyond GitHub's rerun limits, tolerate refused mutations, publish atomically across repositories, reproduce archives across runner-toolchain changes, provide immutable attestations, or use latest-release semantics that GitHub's legacy SemVer selection cannot express.